The rate of diversification of the scientific workforce has been extraordinarily slow despite the efforts of thousands of individuals and large financial investments from government and private agencies over the past 4+ decades. Most research designed to understand and improve this situation has focused on college students and ways to: keep them in STEM majors, increase their sense of fitting-in with STEM fields, and increase their entry to PhD programs. Modest successes have been achieved in the entry of students from underrepresented (UR) ethnic and racial groups into biomedical PhD training. However, their completion of the PhD and success in careers has NOT kept up with even these modest gains. There is a great need for new approaches to improve persistence and success of UR individuals during the PhD and beyond, but very little is known about the true reasons for the fall-off in persistence and success. This proposal would merge two funded research streams targeted at understanding and addressing this problem: 1) a large-scale, empirical study, began in 2008, of experiences and decision-making of biomedical PhD students during their training called The National Longitudinal Study of Young Life Scientists (NLSYLS); 2) translational research, begun in 2011, of a randomized controlled trial of a novel coaching model to augment the idiosyncratic and uneven mentoring students receive, and addressing the isolation and extra challenges faced by young UR scientists - The Academy for Future Science Faculty (Academy). Both studies draw on multiple social science theories to interpret students' experiences during training and design an intervention to improve success. A true understanding of the reasons for the fall-off in success of young UR scientists can only be achieved through longitudinal analysis over time with many students. It also requires much deeper information than can be achieved in surveys. Thus, the NLSYLS employs annual in-depth interviews from the start of the PhD. Analysis of in-depth interviews is extremely complex and time-consuming. To manage and analyze the large volume of interview data, we have developed new approaches to data reduction and display. Interpretation of results also relies on multiple time points. The longitudinal study is just now reaching the point where the first students are graduating and findings can start to be reported. The size, duration and depth of the NLSYLS is beyond anything attempted in the past. Likewise, the Academy is unlike any previous intervention in theoretical underpinnings, design, execution, size and duration. Highly skilled faculty mentors are provided extra training to become Academic Career Coaches. During the Academy, Coaches and students come together for in- person and virtual coaching groups which have been sustained for several years. Coaching is being tested on both beginning and advanced PhD students with concurrent controls. As with the NLSYLS, the Academy experiment is just now to a point where we can begin to see if it is impacting persistence toward academic careers. Integrating these two lines of research with sustained support will lead to extraordinary new insights.